基于自适应Boosting组合模型的空气质量预测  

Air quality prediction based on adaptive boosting combinatorial model

作　　者：徐海峰 黄小莉 张政 Xu Haifeng;Huang Xiaoli;Zhang Zheng(School of Electrical and Electronic Information,Xi Hua University,Chengdu 610000,China)

机构地区：[1]西华大学电气与电子信息学院,四川成都610000

出　　处：《网络安全与数据治理》2022年第12期84-89,共6页CYBER SECURITY AND DATA GOVERNANCE

基　　金：四川省科学技术厅应用基金重点项目(2019YJ0455)。

摘　　要：针对当前空气质量预测模型存在较大误差以及单一模型在不同方面具有一定局限性,导致预测误差较大的问题,提出一种自适应Boosting组合模型。利用误差平方和倒数法、简单加权平均法等五种方法自适应地赋予三个基础Boosting模型权重,最终的结果按照相应的权重叠加,达到充分发挥每个单一模型的优势,提高预测精度的目的。实验结果表明,误差平方和倒数法组合模型的表现最优,采用误差平方和倒数法计算权重的组合模型的MAE为7.1244,RMSE为9.3671,R2为0.8639,优于其他地权重组合方法以及单一的Boosting模型。该组合模型的应用,为优化空气质量预测系统提供了一个行之有效的方法。Aiming at the large error of the current air quality prediction model and the fact that a single model has certain limitations in different aspects,resulting in large prediction error,an adaptive boosting combination model is proposed.Five methods,including the squared error sum reciprocal method and the simple weighted average method,are used to adaptively assign weights to the three basic boosting models.The final result is superimposed according to the corresponding weights,to give full play to the advantages of each single model and improve the prediction accuracy.The experimental results show that,the combination of squared error and reciprocal method performs optimally,and the weighted combinatorial model using the error squared and reciprocal method was calculated with MAE 7.1244,RMSE 9.3671,and R2 as 0.8639,better than other weighting combination methods and a single boosting model.The application of this combined model provides an effective method for optimizing air quality prediction systems.

关 键 词：空气质量 自适应 Boosting模型组合模型 误差平方和倒数法 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]